GPU-accelerated vector indexing slashes OpenSearch build times from days to hours
Amazon OpenSearch Service now supports GPU‑powered k‑NN indexing via NVIDIA cuVS, turning billion‑vector index builds into a matter of hours. The blog details decoupled GPU workers, conversion to HNSW for CPU queries, a 1‑billion‑vector benchmark, and production best practices—crucial for any AI‑driven search workload scaling.